How to vet an n8n expert
Vet an n8n expert in three steps: read what they have published, pay for a small test task, and inspect the workflow they deliver. The canvas itself tells you most of what you need to know, even if you are not technical.
Step 1: read their public work
Skilled n8n people tend to leave a trail. Ten minutes with it will filter out most weak candidates before you spend money.
- Forum answers. Search the candidate's name on the n8n community forum. Clear, patient answers to other people's problems are a strong signal.
- Templates and tutorials. Open one of their shared workflows or videos. Check whether the logic is explained or just shown.
- Code. For custom nodes or integrations, look for a public repository and see whether it is maintained.
- Directory profile. If they appear in n8n's partner directory, read the stated budgets, languages and availability.
- References. One short call with a past client beats ten written testimonials.
Step 2: run a paid test task
Keep it to two to four hours and pay the person's normal rate. Unpaid tests put off the best candidates. Use sample data that resembles yours, with real quirks left in. Here is a task that works for most businesses.
- A webhook receives a new order or lead as JSON.
- The workflow validates required fields and rejects incomplete records with a clear message.
- It looks up the customer in a second system through an API call.
- It writes the combined record to a spreadsheet or database, without creating duplicates when the same record arrives twice.
- It posts a short summary to a chat channel or email.
- One record in your sample data is deliberately broken. The workflow must not stop silently when it meets it.
Step 3: read the workflow
Ask for a screen share or an export of the finished workflow and compare it against this table.
| Area | Good sign | Warning sign |
|---|---|---|
| Naming | Every node has a name that says what it does | Default names such as HTTP Request 3 |
| Notes | Sticky notes explain each stage and any odd decision | A bare canvas that only its author can follow |
| Errors | An error workflow with the Error Trigger node, plus alerts to a person | No error path at all |
| Duplicates | Running the same input twice gives one record | A second run creates a second row |
| Credentials | Stored in n8n's credential manager | API keys pasted into node fields or code |
| Structure | Repeated logic moved into a sub-workflow | One very long chain with copied sections |
| Data handling | Only the needed fields are passed on | Whole payloads, including personal data, sent to every step |
Questions to ask afterwards
Good candidates name their assumptions and point out weaknesses in their own solution. Be careful with anyone who says the test was easy and nothing could be improved.
- Walk me through what happens when the API in step three is down for an hour.
- What would you change if this had to handle a hundred times the volume?
- Which part of the task was unclear, and what did you assume?
- How would you hand this over to someone on my team?
- What would you have done differently with a full day?
Practical checks before you sign
- Agree who owns the workflows and that the n8n instance stays in your account.
- Confirm the time zone, working hours and how fast they reply to failures.
- Ask what happens during their holidays or illness.
- Put the rate, the billing rhythm and a notice period in writing.
- Limit access to the systems the project needs, and remove it when the work ends.
Common questions
How long should the test task take?
Two to four hours is enough. Longer tests cost more and tell you little extra, and strong freelancers often decline them. The goal is to see habits such as naming, error handling and communication, not a finished product.
I am not technical. Can I still judge the result?
Yes. Ask the candidate to explain the workflow on a call as if to a new colleague. If you can follow what each stage does, where errors go and who gets alerted, the workflow is probably well built. Confusion is a finding in itself.